Each simulation is 600 observations of 5 varaibles. There are 3 clusters (200 observations each). The standard deviation used is about twice that used in the mclust figure (harder to see cluster seperation).
The starting basis initialized to PC1:2 of each model. Radial tours created for the top 3 variables.
Scrucca, Luca, Michael Fop, T. Brendan Murphy, and Adrian E. Raftery. “Mclust 5: Clustering, Classification and Density Estimation Using Gaussian Finite Mixture Models.” The R Journal 8, no. 1 (August 2016): 289-317.